Despite the outstanding advances in Perception and Cognition, robots operating in unstructured environments are still at their infancy when it comes to Action. An effective physical action on the environment can be evaluated in terms of motion agility and endurance. Conventional robots based on rigid-body mechanics exhibit a tradeoff betweem agility and energy consumptions. A redundant number of Degrees of Freedoms (DoFs) enables a variety of movements but requires complex and energy-hungry control algorithms. On the contrary, living organisms are nimble and yet energetically efficient, suggesting that the tradeoff between endurance and agility is a limit of our current robots designs.

The intuition coming from the observation of living organisms is that they heavily rely on the complex physics of body-environment interaction to perform low level control tasks that help save substantial amounts of energy. The passive swing of legs, the recoil of the elastic energy stored in tendons, and even the peculiar shape morphing of fish and birds to harness fluidic drag, are only a few examples of how physical characteristics become paramount to efficiently perform Action.

This observation paved the way to Soft Robotics, where more focus is given on the physical embodiment of the robots, introducing mechanical compliance and underactuation by design, and exploring new materials, shapes and control strategies. Due to the emphasis on the design and the possibility offered by the nonlinear mechanics of soft materials, researchers are going beyond structural compliance and moving elements, and are starting to also embed control schemes through physical mechanisms.

The mechanisms enabling those schemes rely on the complex nonlinear physics of soft materials, from large deformations to reversible snap-through instabilities and responsiveness to other physical stimuli. Such mechanisms are typically networks of active or passive soft elements that nonlinearly interact within each other, so that the overall response of the network corresponds to the desired output (an oscillation or a pre-programmed sequence).

In this lecture I show that when multiple nonlinear soft actuators are interconnected they can also embody the control function, by leveraging the local negative stiffness of the actuators to drive their motion out of phase. This allows soft robots to move in pre-programmed sequence using only a single input.